A test problem generation methodology for nonlinear goal programming

We propose a methodology for generating nonlinear goal programs that are suitable for the testing of algorithms. We restrict attention to the most common variant of the class, the preemptive or lexicographic goal program. Our methodology produces test instances that are accompanied by information on how close any optimal solution would come to satisfying each of the goals. Our technique for constructing each test instance is similar in form to sequential optimization procedures for solving goal programs. The method can incorporate varying degrees of randomization